Class level research on modifying information transmission path of the cooperative network based on the cohesion of class

The traditional and mainstream method to study the software system with the complex network is building a collaborative network of the software system on the class level, which is effective to study the software complexity. In the class collaborative network, if the information is passed from a node, it may pass through the nodes at different probabilities. So, the node can block some information dissemination. Due to this feature, there are some problems, such as traverse path errors, inaccuracy of topological properties, inaccuracy propagation domains and propagation costs. Therefore, this paper explores the attributes that affect the probability that information dissemination based on the comparison between topological property of network class nodes and the structure of class. Besides, this paper defines the cohesion of matrix and designs a method of using the cohesive matrix to eliminate the error traversal path method. The experiment shows that this method is valid and effective to exclude error path and is great significant for the class collaborative network.